Hitherto, various kinds of recognition processing are performed using image data imaged with image sensors. Data flows in a system that performs such recognition processing is broadly divided into a data flow that outputs image data as a display image and a data flow that extracts necessary information from image data and performs recognition processing. In order to display an image, generally, a processing speed of about 30 to 120 fps (frame per sec) is sufficient, but, in order to perform advanced recognition processing, it is insufficient. Then, a vision chip in which a processing element is provided for each pixel of an image sensor has been proposed (for example, refer to Patent Literature 1).